Fault detection of multi-information fusion wind turbine pitch system based on dynamic CNN

As a key control component of wind turbines, the failure of pitch system has become one of the main reasons for unit shutdown. Therefore, reducing the failure rate of the pitch system is very important for the stable operation and safety of the unit. In order to extract the fault characteristics in the fault data of the wind turbine pitch system more effectively, a multi-information fusion fault diagnosis model based on dynamic convolutional neural network – CBAM-BiLSTM model is proposed. Based on the BiLSTM network used to extract the timing fault features, a spatial attention CBAM module is introduced, and the fault features of the wind turbine pitch system at the spatial scale are extracted by dynamic convolution. Due to the small volume, fast calculation speed and high accuracy of dynamic convolution, the fault diagnosis speed and accuracy of this model are also improved.